1. Workflow supports text-to-image generation, just disable the Depth module, and the prompts can be generated using reverse inference or written manually.

2. Based on FLUX text-to-image generation, the Depth ControlNet model is added, allowing reference to the poses of characters in the image, combined with RH official image reverse inference prompt nodes.

3. You can customize the F.1 Lora model, which enables generating portraits based on the pose of the reference image and your preferred Lora. Note that chaining multiple Loras may result in lower pixel quality.

4. Added TTP image enlargement function, doubling the size for improved details.

5. Testing shows that image-to-image generation produces quite good results. You can try drawing more cards, and with the enlargement nodes included, image-to-image generation typically takes about 4 minutes.